Axion Echoes and Dark Matter Signals

• Axion echoes are electromagnetic signals produced when radiation near half the axion mass stimulates the decay of nonrelativistic axion dark matter into two photons, resulting in narrow spectral lines.
• They are observed via both controlled laboratory beams and astrophysical sources, with distinct signatures such as back-light, front-light, and collinear emissions.
• Experiments using radio arrays, like FAST and SKA, exploit the echo power dependence on local axion density and source history to achieve sensitivity beyond current limits.

Searching arXiv for papers on axion echoes and stimulated axion decay. Axion echoes are electromagnetic signals generated when radiation at angular frequency close to half the axion mass stimulates the decay of nonrelativistic axion dark matter or axion-like particles into two photons. In the terrestrial configuration, electromagnetic radiation sent out to space can produce detectable back-scattering as faint electromagnetic radiation traveling in the opposite direction; in astrophysical configurations, bright radio sources generate counterimages or “gegenschein” signals, so that axions effectively behave as imperfect monochromatic mirrors (Arza et al., 2019, Buen-Abad et al., 2021, Sun et al., 2023). The subject developed from a controlled beam proposal to detailed analyses of supernova remnants, pulsars, all-sky radio backgrounds, fine-grained streams, and multiple-ALP scenarios, with the radio band emerging as the primary observational regime for $\mu{\rm eV}$-scale masses (Arza et al., 2021, Todarello et al., 2023).

1. Physical basis and resonance

The underlying interaction is the axion–photon coupling

$$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$

which permits the spontaneous decay $a\to\gamma\gamma$ with width

$$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$

In the presence of an ambient photon occupation number, this decay is Bose-enhanced; in the large-occupation regime relevant for bright radio sources, the stimulated contribution dominates over the spontaneous one (Sun et al., 2021, Todarello et al., 2023).

The kinematic resonance is set by the nonrelativistic nature of halo axions. To leading order in the axion velocity, the outgoing stimulating radiation must satisfy $2\omega_{\rm in}\simeq m_a$, and the echo photon is centered at

$\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$

This is why $m_a\lesssim10^{-5}\,{\rm eV}$ maps naturally into radio frequencies below about $1\,{\rm GHz}$, while higher masses move the signal into the microwave regime (Buen-Abad et al., 2021, Arza et al., 2021).

For a transmitted beam, the returned power takes the characteristic form

$$P_{\rm echo}=\frac{1}{16}\,g_{a\gamma\gamma}^2\,\rho_a\,\frac{dP_0}{d\nu}\,t$$

in the simple parallel-beam description, or equivalently

$$P_{\rm echo}=\frac{1}{16}\,g_{a\gamma\gamma}^2\,\rho_a\,\frac{dP_0}{d\nu}\,\ell$$

when expressed in terms of the path length through the axion medium. In both versions, the signal is proportional to the local axion density and to the spectral power of the stimulating beam at the resonant frequency (Arza et al., 2019, Arza et al., 2022).

2. Kinematics, geometry, and image formation

The simplest picture is a two-photon final state in which one photon is emitted forward, effectively cloning the stimulating photon, while the second is emitted backward as the echo. In the idealized axion rest frame the photons are back-to-back; in the laboratory frame, the axion velocity induces both a Doppler shift and an angular deflection. A convenient parameterization is

$$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$0

with $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$1 and $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$2 the axion velocity components parallel and transverse to the stimulating beam (Arza et al., 2022).

The line is narrow because the halo is nonrelativistic. For a Maxwell–Boltzmann halo, the characteristic bandwidth is proportional to $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$3, with representative expressions

$$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$4

At $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$5 and $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$6, one detailed analysis gives $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$7; astrophysical analyses quote a comparable fractional broadening and angular smearing of order a few arcminutes (Arza et al., 2021, Sun et al., 2021, Sun et al., 2023).

The geometry is richer for astrophysical stimulators. “Anatomy of astrophysical echoes from axion dark matter” distinguishes three signatures: a back-light echo, a front-light echo, and collinear emission. For the back-light configuration, the time delay relative to a direct photon is $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$8, while in the front-light configuration $$\mathcal{L}_{a\gamma\gamma}=-\frac14\,g_{a\gamma\gamma}\,a\,F_{\mu\nu}\tilde F^{\mu\nu} \simeq g_{a\gamma\gamma}\,a\,\mathbf E\!\cdot\!\mathbf B\,,$$9 in the point-source, static limit. The back-light and front-light echoes are spread over arcminute scales and are fixed in time, whereas the collinear component inherits any pulsations of the source (Todarello et al., 2023).

A notable result of the detailed beam analysis is that the divergence of the outgoing beam does not affect the echo signal. In the Green’s-function treatment, the same transverse beam profile enters both the incident and stimulated photon wavefunctions, so paraxial divergences cancel in the returned amplitude; the limiting factor is the axion velocity distribution rather than beam spread (Arza et al., 2021).

3. Astrophysical echo phenomenology

Astrophysical echo searches replace the artificial transmitted beam by an existing source in the sky. For a source of age $a\to\gamma\gamma$0, the echo flux density depends on the entire retarded light curve:

$a\to\gamma\gamma$1

This expression makes explicit that the signal is weighted by the dark-matter density along the echo leg and by the source luminosity at earlier times. The physical consequence is central: the strength of the so-called axion “echo” signal depends on the entire history of the source and could still be strong from sources that are dim today but had a large flux density in the past (Buen-Abad et al., 2021).

Supernova remnants are the canonical example. Their radio-bright phases are long enough that the early, brighter stages can dominate the echo integral. This motivates searches not only around cataloged remnants, but also around old “ghost” SNRs which were very bright in the past but are now so dim that they haven’t been observed. The same logic underlies the “supernova graveyard” picture: a faint present-day remnant can remain an effective stimulator because the echo encodes its historical luminosity rather than only its current radio flux (Buen-Abad et al., 2021).

The source catalog can be broadened well beyond SNRs. An all-sky analysis uses extragalactic radio point sources, Galactic SNRs, and Galactic synchrotron radiation as sources of stimulating radiation. In that treatment, the aggregate signal strength is not significantly affected by unknown properties of individual sources, because those properties are sampled from an empirical distribution to generate an ensemble of realizations for the all-sky signal template (Sun et al., 2023).

Pulsars supply a complementary geometry. For Galactic pulsars, the three associated signatures—two echoes and one collinear emission—can be simultaneously detected in principle. The joint pattern is highly constraining: the back-light echo and front-light echo are orthogonally polarized relative to the original signal and are unpulsed, whereas the collinear emission is pulsed and appears exactly at the source position (Todarello et al., 2023).

4. Experiments, facilities, and quoted sensitivities

Quoted sensitivities span dedicated beam experiments, targeted observations of individual astrophysical sources, and survey-mode analyses with existing radio arrays (Arza et al., 2019, Sun et al., 2021, Buen-Abad et al., 2021, Arza et al., 2023, Sun et al., 2023, Todarello et al., 2023).

Configuration	Window	Quoted reach
Dedicated outgoing-beam echo search	$a\to\gamma\gamma$2–$a\to\gamma\gamma$3	caustic-ring “Big Flow”: $a\to\gamma\gamma$4; isothermal model: $a\to\gamma\gamma$5 near $a\to\gamma\gamma$6
FAST targeted SNR gegenschein search	$a\to\gamma\gamma$7–$a\to\gamma\gamma$8	$a\to\gamma\gamma$9 excluded with $$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$0 per source
SKA1 search with known SNRs	$$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$1–$$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$2	$$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$3–$$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$4
21CMA with a $$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$5 emitter	$$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$6–$$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$7	standard layout: $$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$8; optimized layout: $$\Gamma_{a\to\gamma\gamma}^{(0)}=\frac{g_{a\gamma\gamma}^2\,m_a^3}{64\pi}\,.$$9
CHIME, HERA, CHORD, HIRAX, BURSTT all-sky surveys	$2\omega_{\rm in}\simeq m_a$0–$2\omega_{\rm in}\simeq m_a$1, facility dependent	$2\omega_{\rm in}\simeq m_a$2
Pulsars with SKA-1 or LOFAR 2.0	$2\omega_{\rm in}\simeq m_a$3–$2\omega_{\rm in}\simeq m_a$4	$2\omega_{\rm in}\simeq m_a$5 in $2\omega_{\rm in}\simeq m_a$6; full pulsar survey could reach $2\omega_{\rm in}\simeq m_a$7

The observing strategies differ, but the statistical structure is similar. Several analyses adopt a radiometer or Dicke-type signal-to-noise estimate, with the optimal bandwidth chosen to match the Doppler-broadened line. Representative choices are $2\omega_{\rm in}\simeq m_a$8 with $2\omega_{\rm in}\simeq m_a$9 in the SNR forecasts and $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$0 with $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$1 in the all-sky survey forecasts (Buen-Abad et al., 2021, Sun et al., 2023).

A recurring feature is complementarity with helioscopes and haloscopes. The SNR studies emphasize that SKA1 and FAST can reach significantly below present CAST limits in relevant mass windows, while the all-sky interferometer study stresses that these searches can run simultaneously with 21 cm cosmology and other survey objectives, requiring no new hardware (Buen-Abad et al., 2021, Sun et al., 2023).

5. Dependence on phase-space structure and systematics

The echo signal is unusually sensitive to the local axion phase-space distribution. In the original beam proposal, the contrast between the isothermal and caustic-ring models is substantial: the caustic “Big Flow” case combines larger density, narrower linewidth, and smaller transverse velocity, whereas the isothermal model requires much larger outgoing energy per octave and yields a weaker coupling reach (Arza et al., 2019).

Fine-grained streams sharpen this dependence further. In the Standard Halo Model one has $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$2, whereas individual streams are modeled with intrinsic dispersion $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$3. Because the collected power scales as $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$4, the narrower stream line can improve the sensitivity to $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$5 by up to two orders of magnitude relative to the isotropic SHM. Detailed projections for the Green Bank Telescope and SKA are reported as reaching well into QCD-axion parameter space, and a summary statement gives $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$6 in the mass range $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$7–$\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$8 with existing facilities (Arza et al., 2022).

Gravitationally focused “hairs” provide a different nonstandard target. In that scenario the Earth focuses a fine-grained stream into a filament with amplification factor $\nu_{\rm echo}=\frac{m_a}{4\pi}\,.$9, but the small diameter and the Earth’s rotation limit the observing time to $m_a\lesssim10^{-5}\,{\rm eV}$0. After including the hair encounter rate, the net gain in coupling reach is quoted as up to an order of magnitude compared to the SHM (Arza et al., 2022).

Several potential misconceptions are explicitly addressed in the literature. Beam divergence is not the main degradation mechanism; the returned amplitude retraces the incident wavefront up to paraxial order, and the limiting effects are the velocity distribution and the detector bandwidth (Arza et al., 2021). Likewise, in the all-sky astrophysical treatment, the aggregate signal strength is not significantly affected by unknown properties of individual sources, because the signal template is built from an ensemble of realizations rather than a single deterministic catalog model (Sun et al., 2023).

6. Extensions beyond the single-ALP radio echo

A recent extension studies multiple ALPs coupled to the photon. In the coherent case, if $m_a\lesssim10^{-5}\,{\rm eV}$1 ALPs oscillate in phase with identical masses and couplings, the echo power scales linearly with $m_a\lesssim10^{-5}\,{\rm eV}$2,

$m_a\lesssim10^{-5}\,{\rm eV}$3

and the reachable coupling bound improves by $m_a\lesssim10^{-5}\,{\rm eV}$4. Small mass splittings can further increase the amplification, even for $m_a\lesssim10^{-5}\,{\rm eV}$5. In the incoherent case, random phases suppress the enhancement: the random-phase average replaces coherent addition by a root-mean-square effective amplitude, and the observable signal becomes akin to or weaker than the single-ALP case (Haque et al., 22 Jul 2025).

A distinct usage of the echo concept appears in core-collapse supernovae at much higher masses. In that scenario, ALPs produced in the proto-neutron-star core decay in the stellar envelope, generating positrons that thermalize and annihilate at rest, producing a characteristic echo of $m_a\lesssim10^{-5}\,{\rm eV}$6 gamma rays. For SN 1987A, Pioneer Venus Orbiter non-observation yields

$m_a\lesssim10^{-5}\,{\rm eV}$7

while forecasts for a Galactic supernova give $m_a\lesssim10^{-5}\,{\rm eV}$8 for COSI and $m_a\lesssim10^{-5}\,{\rm eV}$9 for AMEGO near a few hundred MeV (Chauhan et al., 18 Nov 2025).

Taken together, these developments place axion echoes at the intersection of radio astronomy, dark-matter phase-space modeling, and multimessenger astrophysics. In the standard radio formulation, the defining observable remains a narrow spectral line at $1\,{\rm GHz}$0, with bandwidth and angular structure fixed by the axion velocity distribution and with amplitude controlled either by the transmitted power of a laboratory beam or by the time history of an astrophysical radio source.